kreuzberg 3.16.0__py3-none-any.whl → 3.17.0__py3-none-any.whl

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (90) hide show
  1. kreuzberg/__init__.py +2 -0
  2. kreuzberg/_config.py +8 -9
  3. kreuzberg/_extractors/_base.py +0 -46
  4. kreuzberg/_extractors/_html.py +1 -1
  5. kreuzberg/_extractors/_pandoc.py +2 -2
  6. kreuzberg/_extractors/_pdf.py +4 -4
  7. kreuzberg/_gmft.py +2 -2
  8. kreuzberg/_mcp/server.py +1 -1
  9. kreuzberg/_mime_types.py +1 -1
  10. kreuzberg/_ocr/_easyocr.py +4 -9
  11. kreuzberg/_ocr/_paddleocr.py +1 -1
  12. kreuzberg/_ocr/_tesseract.py +15 -25
  13. kreuzberg/_token_reduction/__init__.py +11 -0
  14. kreuzberg/_token_reduction/_reducer.py +439 -0
  15. kreuzberg/_token_reduction/_stopwords.py +116 -0
  16. kreuzberg/_token_reduction/stopwords/af_stopwords.json +53 -0
  17. kreuzberg/_token_reduction/stopwords/ar_stopwords.json +482 -0
  18. kreuzberg/_token_reduction/stopwords/bg_stopwords.json +261 -0
  19. kreuzberg/_token_reduction/stopwords/bn_stopwords.json +400 -0
  20. kreuzberg/_token_reduction/stopwords/br_stopwords.json +1205 -0
  21. kreuzberg/_token_reduction/stopwords/ca_stopwords.json +280 -0
  22. kreuzberg/_token_reduction/stopwords/cs_stopwords.json +425 -0
  23. kreuzberg/_token_reduction/stopwords/da_stopwords.json +172 -0
  24. kreuzberg/_token_reduction/stopwords/de_stopwords.json +622 -0
  25. kreuzberg/_token_reduction/stopwords/el_stopwords.json +849 -0
  26. kreuzberg/_token_reduction/stopwords/en_stopwords.json +1300 -0
  27. kreuzberg/_token_reduction/stopwords/eo_stopwords.json +175 -0
  28. kreuzberg/_token_reduction/stopwords/es_stopwords.json +734 -0
  29. kreuzberg/_token_reduction/stopwords/et_stopwords.json +37 -0
  30. kreuzberg/_token_reduction/stopwords/eu_stopwords.json +100 -0
  31. kreuzberg/_token_reduction/stopwords/fa_stopwords.json +801 -0
  32. kreuzberg/_token_reduction/stopwords/fi_stopwords.json +849 -0
  33. kreuzberg/_token_reduction/stopwords/fr_stopwords.json +693 -0
  34. kreuzberg/_token_reduction/stopwords/ga_stopwords.json +111 -0
  35. kreuzberg/_token_reduction/stopwords/gl_stopwords.json +162 -0
  36. kreuzberg/_token_reduction/stopwords/gu_stopwords.json +226 -0
  37. kreuzberg/_token_reduction/stopwords/ha_stopwords.json +41 -0
  38. kreuzberg/_token_reduction/stopwords/he_stopwords.json +196 -0
  39. kreuzberg/_token_reduction/stopwords/hi_stopwords.json +227 -0
  40. kreuzberg/_token_reduction/stopwords/hr_stopwords.json +181 -0
  41. kreuzberg/_token_reduction/stopwords/hu_stopwords.json +791 -0
  42. kreuzberg/_token_reduction/stopwords/hy_stopwords.json +47 -0
  43. kreuzberg/_token_reduction/stopwords/id_stopwords.json +760 -0
  44. kreuzberg/_token_reduction/stopwords/it_stopwords.json +634 -0
  45. kreuzberg/_token_reduction/stopwords/ja_stopwords.json +136 -0
  46. kreuzberg/_token_reduction/stopwords/kn_stopwords.json +84 -0
  47. kreuzberg/_token_reduction/stopwords/ko_stopwords.json +681 -0
  48. kreuzberg/_token_reduction/stopwords/ku_stopwords.json +64 -0
  49. kreuzberg/_token_reduction/stopwords/la_stopwords.json +51 -0
  50. kreuzberg/_token_reduction/stopwords/lt_stopwords.json +476 -0
  51. kreuzberg/_token_reduction/stopwords/lv_stopwords.json +163 -0
  52. kreuzberg/_token_reduction/stopwords/ml_stopwords.json +11 -0
  53. kreuzberg/_token_reduction/stopwords/mr_stopwords.json +101 -0
  54. kreuzberg/_token_reduction/stopwords/ms_stopwords.json +477 -0
  55. kreuzberg/_token_reduction/stopwords/ne_stopwords.json +490 -0
  56. kreuzberg/_token_reduction/stopwords/nl_stopwords.json +415 -0
  57. kreuzberg/_token_reduction/stopwords/no_stopwords.json +223 -0
  58. kreuzberg/_token_reduction/stopwords/pl_stopwords.json +331 -0
  59. kreuzberg/_token_reduction/stopwords/pt_stopwords.json +562 -0
  60. kreuzberg/_token_reduction/stopwords/ro_stopwords.json +436 -0
  61. kreuzberg/_token_reduction/stopwords/ru_stopwords.json +561 -0
  62. kreuzberg/_token_reduction/stopwords/si_stopwords.json +193 -0
  63. kreuzberg/_token_reduction/stopwords/sk_stopwords.json +420 -0
  64. kreuzberg/_token_reduction/stopwords/sl_stopwords.json +448 -0
  65. kreuzberg/_token_reduction/stopwords/so_stopwords.json +32 -0
  66. kreuzberg/_token_reduction/stopwords/st_stopwords.json +33 -0
  67. kreuzberg/_token_reduction/stopwords/sv_stopwords.json +420 -0
  68. kreuzberg/_token_reduction/stopwords/sw_stopwords.json +76 -0
  69. kreuzberg/_token_reduction/stopwords/ta_stopwords.json +129 -0
  70. kreuzberg/_token_reduction/stopwords/te_stopwords.json +54 -0
  71. kreuzberg/_token_reduction/stopwords/th_stopwords.json +118 -0
  72. kreuzberg/_token_reduction/stopwords/tl_stopwords.json +149 -0
  73. kreuzberg/_token_reduction/stopwords/tr_stopwords.json +506 -0
  74. kreuzberg/_token_reduction/stopwords/uk_stopwords.json +75 -0
  75. kreuzberg/_token_reduction/stopwords/ur_stopwords.json +519 -0
  76. kreuzberg/_token_reduction/stopwords/vi_stopwords.json +647 -0
  77. kreuzberg/_token_reduction/stopwords/yo_stopwords.json +62 -0
  78. kreuzberg/_token_reduction/stopwords/zh_stopwords.json +796 -0
  79. kreuzberg/_token_reduction/stopwords/zu_stopwords.json +31 -0
  80. kreuzberg/_types.py +35 -3
  81. kreuzberg/_utils/_image_preprocessing.py +1 -1
  82. kreuzberg/_utils/_ref.py +14 -6
  83. kreuzberg/exceptions.py +0 -1
  84. kreuzberg/extraction.py +25 -9
  85. {kreuzberg-3.16.0.dist-info → kreuzberg-3.17.0.dist-info}/METADATA +4 -3
  86. kreuzberg-3.17.0.dist-info/RECORD +128 -0
  87. kreuzberg-3.16.0.dist-info/RECORD +0 -61
  88. {kreuzberg-3.16.0.dist-info → kreuzberg-3.17.0.dist-info}/WHEEL +0 -0
  89. {kreuzberg-3.16.0.dist-info → kreuzberg-3.17.0.dist-info}/entry_points.txt +0 -0
  90. {kreuzberg-3.16.0.dist-info → kreuzberg-3.17.0.dist-info}/licenses/LICENSE +0 -0
kreuzberg/_token_reduction/stopwords/et_stopwords.json ADDED
@@ -0,0 +1,37 @@
1
+ [
2
+ "aga",
3
+ "ei",
4
+ "et",
5
+ "ja",
6
+ "jah",
7
+ "kas",
8
+ "kui",
9
+ "kõik",
10
+ "ma",
11
+ "me",
12
+ "mida",
13
+ "midagi",
14
+ "mind",
15
+ "minu",
16
+ "mis",
17
+ "mu",
18
+ "mul",
19
+ "mulle",
20
+ "nad",
21
+ "nii",
22
+ "oled",
23
+ "olen",
24
+ "oli",
25
+ "oma",
26
+ "on",
27
+ "pole",
28
+ "sa",
29
+ "seda",
30
+ "see",
31
+ "selle",
32
+ "siin",
33
+ "siis",
34
+ "ta",
35
+ "te",
36
+ "ära"
37
+ ]
kreuzberg/_token_reduction/stopwords/eu_stopwords.json ADDED
@@ -0,0 +1,100 @@
1
+ [
2
+ "al",
3
+ "anitz",
4
+ "arabera",
5
+ "asko",
6
+ "baina",
7
+ "bat",
8
+ "batean",
9
+ "batek",
10
+ "bati",
11
+ "batzuei",
12
+ "batzuek",
13
+ "batzuetan",
14
+ "batzuk",
15
+ "bera",
16
+ "beraiek",
17
+ "berau",
18
+ "berauek",
19
+ "bere",
20
+ "berori",
21
+ "beroriek",
22
+ "beste",
23
+ "bezala",
24
+ "da",
25
+ "dago",
26
+ "dira",
27
+ "ditu",
28
+ "du",
29
+ "dute",
30
+ "edo",
31
+ "egin",
32
+ "ere",
33
+ "eta",
34
+ "eurak",
35
+ "ez",
36
+ "gainera",
37
+ "gu",
38
+ "gutxi",
39
+ "guzti",
40
+ "haiei",
41
+ "haiek",
42
+ "haietan",
43
+ "hainbeste",
44
+ "hala",
45
+ "han",
46
+ "handik",
47
+ "hango",
48
+ "hara",
49
+ "hari",
50
+ "hark",
51
+ "hartan",
52
+ "hau",
53
+ "hauei",
54
+ "hauek",
55
+ "hauetan",
56
+ "hemen",
57
+ "hemendik",
58
+ "hemengo",
59
+ "hi",
60
+ "hona",
61
+ "honek",
62
+ "honela",
63
+ "honetan",
64
+ "honi",
65
+ "hor",
66
+ "hori",
67
+ "horiei",
68
+ "horiek",
69
+ "horietan",
70
+ "horko",
71
+ "horra",
72
+ "horrek",
73
+ "horrela",
74
+ "horretan",
75
+ "horri",
76
+ "hortik",
77
+ "hura",
78
+ "izan",
79
+ "ni",
80
+ "noiz",
81
+ "nola",
82
+ "non",
83
+ "nondik",
84
+ "nongo",
85
+ "nor",
86
+ "nora",
87
+ "ze",
88
+ "zein",
89
+ "zen",
90
+ "zenbait",
91
+ "zenbat",
92
+ "zer",
93
+ "zergatik",
94
+ "ziren",
95
+ "zituen",
96
+ "zu",
97
+ "zuek",
98
+ "zuen",
99
+ "zuten"
100
+ ]